Thresholds for the Effects of Pesticides on Invertebrate Communitiesand Leaf Breakdown in Stream EcosystemsRalf B. Schafer,*,† Peter Carsten von der Ohe,‡ Jes Rasmussen,§ Ben J. Kefford,∥ Mikhail A. Beketov,⊥

Ralf Schulz,† and Matthias Liess⊥

†Institute for Environmental Sciences, University Koblenz-Landau, Campus Landau, Fortstrasse 7, 76829 Landau, Germany‡UFZ, Helmholtz Centre for Environmental Research, Department Effect-Directed Analysis, Permoserstrasse 15, 04318 Leipzig,Germany§Faculty of Science and Technology, University of Aarhus, Vejlsøvej 25, 8600 Aarhus, Denmark∥Centre for Environmental Sustainability, School of the Environment, University of Technology Sydney, Broadway NSW 2007,Australia⊥UFZ, Helmholtz Centre for Environmental Research, Department System-Ecotoxicology, Permoserstrasse 15, 04318 Leipzig,Germany

*S Supporting Information

ABSTRACT: We compiled data from eight field studies conductedbetween 1998 and 2010 in Europe, Siberia, and Australia to derivethresholds for the effects of pesticides on macroinvertebratecommunities and the ecosystem function leaf breakdown. Dose−response models for the relationship of pesticide toxicity with theabundance of sensitive macroinvertebrate taxa showed significantdifferences to reference sites at 1/1000 to 1/10 000 of the medianacute effect concentration (EC50) for Daphnia magna, depending onthe model specification and whether forested upstream sections werepresent. Hence, the analysis revealed effects well below the thresholdof 1/100 of the EC50 for D. magna incorporated in the EuropeanUnion Uniform Principles (UP) for registration of single pesticides.Moreover, the abundances of sensitive macroinvertebrates in thecommunities were reduced by 27% to 61% at concentrations relatedto 1/100 of the EC50 for D. magna. The invertebrate leaf breakdown rate was positively linearly related to the abundance ofpesticide-sensitive macroinvertebrate species in the communities, though only for two of the three countries examined. We arguethat the low effect thresholds observed were not mainly because of an underestimation of field exposure or confounding factors.From the results gathered we derive that the UP threshold for single pesticides based on D. magna is not protective for fieldcommunities subject to multiple stressors, pesticide mixtures, and repeated exposures and that risk mitigation measures, such asforested landscape patches, can alleviate effects of pesticides.

■ INTRODUCTIONFreshwater ecosystems are among the most threatenedecosystems in terms of species extinctions and losses inecosystem services. One of the major stressors for theseecosystems are pesticides, which are introduced via point andnonpoint sources.1 An efficient protection of freshwaterecosystems requires the determination of a reliable thresholdvalue for the effects of pesticides. For example, the UniformPrinciples (UP) of the European Union (EU) state that for asingle pesticide “no authorization shall be granted if thetoxicity/exposure ratio for fish and Daphnia is less than 100 foracute exposure [...]”.2 A review of mesocosm studies on theeffects of single insecticides (carbamates, organophosphates,and pyrethroids) by Wijngaarden et al.3 suggested that thissafety factor would be protective. They reported that insecticideconcentrations below the above-mentioned toxicity/exposure

ratio of 100 that relates to a concentration of 1/100 of themedian effect concentration (EC50) for Daphnia magna areunlikely to cause notable effects.3 However, the joint effects ofmultiple stressors, including mixtures of pesticides4 are rarelyconsidered in mesocosm studies,5 though they may influenceeffect thresholds.6 Indeed, a field study conducted in 20agricultural streams showed a significant change in communitystructure already at an acute toxicity/exposure ratio for D.magna in the range of 100−1000 for the most toxiccompound.7

Received: November 8, 2011Revised: March 10, 2012Accepted: March 28, 2012Published: March 28, 2012

Article

pubs.acs.org/est

© 2012 American Chemical Society 5134 dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−5142

Beside effects on the structure of freshwater communities,pesticides can impede important ecosystem functions such asleaf breakdown8 that represents the main energy source infreshwater ecosystems beside gross primary production.9

However, to which extent effects on biota propagate to effectson ecosystem functions has been ranked as one of the mostimportant research questions for the conservation of biologicaldiversity.10 Theoretically, the biotic community and ecosystemfunctions can be linked in four ways.11 First, in a near linearway implying effects on biota would lead to a similar decline inecosystem functions. Second, there may be functionalredundancy in the community and no effects on ecosystemfunctions would occur up to certain thresholds. Third, the lossof most species may be compensated whereas the loss of a fewso-called keystone species or ecosystem engineers would resultin changes in ecosystem functions. Thus the effects depend onthe identity of the species lost. Fourth, a chemical may alter thefunctional capacity of species and hence affect ecosystemfunctioning without alteration of the community.12 It isunknown which of these models applies for the relationshipbetween effects of pesticides on biota and on ecosystemfunctions and whether this relationship would be universal.Species traits have been suggested as a stressor-specific tool

in ecological risk assessment13,14 as they allow for a mechanisticlink between stressors and communities, even under conditionsof multiple stressors.15,16 The SPEcies At Risk (SPEAR)indicator for pesticides7 relies on species traits to calculate thefraction of pesticide-sensitive species in macroinvertebratecommunities. The SPEAR index has been successfully linkedto pesticide toxicity and the leaf breakdown rate in field studies,while being generally discriminative toward co-occurringstressors in agricultural regions,7,14,17−19 as well as applicableover different biogeographical regions.17−19 The latter isespecially important because it enables the meta-analysis ofstudies from different regions.In this study, we determined thresholds for the effects of

pesticides on freshwater ecosystems from a meta-analysis offield studies. Therefore, we compiled data from various fieldstudies in different regions on the effects of pesticides onfreshwater macroinvertebrate communities as detected usingthe SPEAR approach as well as on the ecosystem function leaf

breakdown. Macroinvertebrate communities were selected asstructural end point since (a) they belong to the most sensitivegroup of organisms to pesticides in freshwater communities,(b) trait-based approaches in freshwater ecology are mostadvanced for macroinvertebrates, and (c) there is a paucity offield studies on the effects of pesticides on other groups ofbiota.20 In addition, we examined how effects on the structure,in terms of the fraction of pesticide-sensitive species in thecommunities, are related to effects on the important ecosystemfunction of leaf breakdown.20

■ EXPERIMENTAL SECTION

Selection and Description of Field Studies. Thefollowing inclusion criteria were used for the selection of fieldstudies on the effects of pesticides: (1) at least 5 differentstreams monitored, (2) selection of pesticides for chemicalanalysis that are most likely to represent a risk to macro-invertebrates in the respective region based on recommendedpesticide use information for the respective year (and region)sampled, available toxicity data for D. magna or results fromprevious monitoring programs [see 7, 17, 19] and (3) theSPEAR values or leaf breakdown rates reported. In addition, weincluded reference sites from studies, where SPEAR values werereported (Table 1). We focused on studies reporting the trait-based SPEAR indicator, because in contrast to taxonomical datathis indicator has been demonstrated to be applicable overdifferent biogeographical regions.17,18 However, we are notaware of other studies that met the first two criteria andpresented macroinvertebrate community data to assess theeffects of pesticides. For example, one study encompassing 29different streams21 was not included since only sedimentconcentrations for a limited set of pesticides were reported andthe total sediment concentration of the monitored pesticideswas used as a proxy for nonmonitored pesticide concentrations.Overall, 8 studies conducted between 1998 and 2010 (studyduration between 2.5 and 36 months) with a total of 111 siteswere included in the present study, of which 6 studies wereconducted in different regions of Europe, and a study in each ofAustralia and Siberia (Table 1). Except for reference sites whichwere predominantly located in forested areas, the sites in thestudies were located in agricultural areas. The sites were

Table 1. Study Regions with Number (No.) of Sites, Biological End Points Reported, Number of Pesticides Measured andRange of Toxic Units (TU) Included in This Study

regionno.sites biological end points reported

pesticide monitoringmethodsd

no. of pesticidesmeasured

lowest log TUreported

highest log TUreported ref

South Finland 13 SPEARabundance, SPEARPPM abundance andinvertebrate leaf breakdown

PS and SPS 10 −5 −4.3 17

Brittany, France 16 SPEARabundance, SPEARPPM abundance andinvertebrate leaf breakdown

EWS, PS, and SPS 10 −5 −0.4 17

CentralGermany

20 SPEARabundance 2 EWS 21 −5 −0.7 7

Victoria,Australia

24 SPEARpesticidesb and invertebrate leaf

breakdownGWS, PS, andsediment

97 −3.5 −0.2 8, 19

Island Funen,Denmark

14 SPEARpesticidesb and invertebrate leaf

breakdownEWS, GWS, andsediment

31 −6.6 −1.7 22,23

Flanders,Belgium

7 SPEAR[%]b 0e 18

North Germany 11 SPEAR[%]b 0e 18Siberia, Russia 6a SPEARorganic

c 0e 24

aOnly reference sites included (sites 3, 5, 6, 7, 9, and 10 of original publication24). bReported indicator values were calculated according to Liess andvon der Ohe.7 For this study, the SPEARpesticides values were calculated from the original data as described below. cFor this study, the SPEARpesticidesindicator was calculated from the original data as described below. dPS = passive sampling, SPS = suspended particle sampling, EWS = event-drivenwater sampling, GWS = grab water sampling eSites were considered as having no pesticide contamination. See references for details.

Environmental Science & Technology Article

dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−51425135

selected to not receive discharge from wastewater treatmentplants, industrial facilities or mines in order to exclude the inputof toxicants other than pesticides. The pesticide monitoring wasadjusted (1) to capture episodic runoff events and (2) to theproperties of the pesticides selected for chemical analysis in theparticular study. The selected pesticides varied among the studyregions due to differences in crops, pests and authorizedpesticides (Table 1).Data preparation. Environmental concentrations of

pesticides were scaled to acute effects of D. magna calculatingToxic Units (TU)25 by dividing the compound concentrationwith the respective 48-h median effect concentration (EC50)for D. magna. Dose−response modeling was used to evaluatethe relationship between TU and SPEAR, which representpesticide toxicity and community change, respectively. TheTUs used here were given as the maximum TUs of allpesticides across samples per site in the original studies andwere reported to have similar explanatory power for biotic endpoints as the sum of TUs of all pesticides across each or allsamples per site.7,19,26 The maximum TU represents thesimplest approach because the estimated pesticide toxicity reliessolely on the most toxic pesticide concentration observed persite, whereas all pesticide concentrations per sample or sitecontribute to the calculation of the sum of TU. Carbamate andorganophosphate insecticides and several fungicides werepredominantly responsible for the maximum TU in the sites(Supporting Information Table S1).We used the modified version of the original SPEAR

indicator7 as described in Schafer et al.17 (therein referred to asSPEARPM abundance) to compare the indicator values betweendifferent biogeographical regions. For terminological clarity, werefer to this indicator as SPEARpesticides in the following assuggested by Beketov et al.27 For sites for which this version ofthe SPEAR indicator was not reported (Table 1), we calculatedthe indicator according to:

=∑ +∑ +

=

=

x y

xSPEAR

log( 1)

log( 1)in

i

in

ipesticides

1

1

where n is the number of taxa observed in a sampling site, xi isthe abundance of taxon i and y is 1 if taxon i is classified asSpecies At Risk (SPEAR) regarding the traits “physiologicalsensitivity” and “dispersal capacity”, otherwise 0. The trait dataused were derived from the database associated with theSPEAR online calculator (http://www.systemecology.eu/SPEAR/index.php).To characterize the propagation of effects from pesticide-

driven structural changes to ecosystem functions, the responseof the invertebrate-driven leaf breakdown rate (kinvertebrate)

28 tochanges in SPEARpesticides was investigated. The leaf breakdownrate is not stressor-specific and hence responds to differentenvironmental conditions.29 In contrast to species traits,30

kinvertebrate varies over biogeographical regions and would not beexpected to be similar across sites without pesticidecontamination because of the influence of other environmentalgradients.31 Indeed, no relevant pesticide toxicity and onlyminor variation of SPEARpesticides was detected in the sites fromSouth Finland, whereas the invertebrate leaf breakdown ratevaried strongly between sites in response to temperature.17

Since the aim was to examine the link between pesticide-drivencommunity change and invertebrate leaf breakdown rate, thesesites were not considered for further analysis. The invertebrateleaf breakdown rates from the French, Danish and Australian

streams (Table 1) were not comparable as leafs of different treespecies were employed. Therefore, the % change in the leafbreakdown rate kinvertebrate was calculated for each data set bydividing all values for kinvertebrate by the maximum value obtainedfrom individually fitted models for kinvertebrate as explained bySPEARpesticides (see below). We used the maximum value fromthe models fitted with all data points instead of the maximumvalue for kinvertebrate from the respective raw data in order toavoid undue influence of a single data point.

Data Analysis. Before analysis the data were divided intosites with and without forested upstream sections, as defined inthe original publications, and analyzed separately (Table 1).This was done because the presence of forested upstreamsections was demonstrated to alleviate the effects of pesticideson the macroinvertebrate community as indicated bySPEARpesticides.

7,17,26 S-shaped dose−response curves with TUas concentration and SPEARpesticides as response variable werecomputed using two-parameter log−logistic, Weibull I andWeibull II models with the upper limit fixed to the arithmeticmean of the SPEARpesticides values for reference sites and thelower limit fixed to 0,32 representing the lowest possibleindicator value. In addition, linear, quadratic, and cubicregression models were computed to check for the fit ofmore parsimonious models. The best-fit model among the s-shaped dose response models and the polynomial regressionmodels was selected using the Bayesian Information Criterion(BIC). If a s-shaped dose−response curve represented the best-fit model, we calculated the effect concentration (EC) for thepercentage (p) of reduction in SPEARpesticides for p = 10, 50 and90%. In addition, the p value was computed for the EC relatedto the log TU of −2 that is equal to the safety factor of 100,employed in the UP of the EU. Moreover, we derived thelowest concentration at which significant differences (α = 0.05)to reference sites occur in the best-fit dose−response modelusing 95% confidence intervals. Technically, we determined thelowest concentration for which the 95% confidence interval ofthe fitted model did not overlap with the 95% confidenceinterval for the reference sites. Several of the original studiesassigned log TUs of either −5 or −4 to sites where nopesticides were found assuming that this would represent theminimum log TU for which no pesticide effects would occur(Table 2). Since the minimum TU influences the dose−

response modeling, all analyses were conducted for bothminimum reported TUs, that is, we assigned either a log TU of−5 or −4 to all sites with no pesticide detections. To confirmthe results of the dose−response modeling, analysis of variance(ANOVA) with a priori treatment contrasts was used toidentify significant differences in SPEARpesticides values betweenreference sites and groups of contaminated sites in terms ofTU. The class boundaries of log TU ≤ −3.5 (reference sites),

Table 2. Estimated Effect Concentrations (EC) in Terms oflog TU for p = 10%, 50%, and 90% Reduction inSPEARpesticides for the Models with and without ForestedUpstream Sections (FUS and WFUS, Respectively) and witha Minimum log TU of −5 (low min.) or −4 (high min.)

estimated EC (in log TU)

p FUS low min. FUS high min. WFUS low min. WFUS high min.

10 −3.6 −2.9 −4.2 −3.550 −1.7 −1.4 −2.5 −2.190 −0.4 −0.4 −0.6 −0.6

Environmental Science & Technology Article

dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−51425136

−3.5 < log TU ≤ −2.5 (lightly contaminated sites), −2.5 < logTU ≤ −1.5 (moderately contaminated sites) and log TU >−1.5 (highly contaminated sites) were selected in order to havesimilar class widths and at least 5 observations in each class. Asimilar classification was used in the study of Schafer et al.17

For the examination of the propagation of effects ofpesticides on ecosystem functions, the relationship betweenthe invertebrate-driven leaf breakdown rate (% kinvertebrate) andSPEARpesticides was modeled using s-shaped dose−responsemodels and polynomial regression models as described above,except that three-parameter s-shaped dose response modelswere fitted since no upper limit could be fixed. First, themodeling was done separately for each country, because in theoriginal studies this relationship had only been examined forthe sites from Brittany, France. Subsequently, the data weremodeled jointly. For best-fit linear regression models,significance of the slope was tested with the t test and in thecase of data from different countries, analysis of covariance(ANCOVA) was used to detect significant differences betweenslopes and intercepts from the countries. ANOVA, ANCOVAand linear regression models were checked for normaldistribution of residuals, homoscedasticity and unusualobservations.33 All computations and graphics were createdwith the free and open source software R (version 2.13.1 forMac OS X, 10.6.8)34 including supplemental packages such as“drc” for dose−response modeling.35

■ RESULTSWeibull I and II as well as log−logistic models were identifiedas best-fit dose−response models for TU and SPEARpesticides(Supporting Information Table S2). The estimated EC10 for thedifferent models ranged from a log TU of −2.9 to a log TU of−4.2 depending on a) the availability of forested upstreamsections and b) which minimum TU was assigned (Table 2).The estimated EC90 were identical for the two minimum TUs(Table 2). For an EC related to a log TU of −2, the fraction ofspecies at risk in the communities in terms of abundance wasreduced by 27% and 41% in sites with forested upstreamsections for models with a minimum log TU of −4 and −5,respectively (Figure 1, Figure S1). In sites without forestedupstream sections, this EC corresponded to 54% and 61%reduction in SPEARpesticides, respectively (Figure 1, Figure S1).Significant differences (nonoverlapping 95% confidence inter-

vals) to reference sites were observed for log TUs ≥ −3 and−3.5 in sites with forested upstream sections and for log TUs ≥−3 and −3.6 in sites without forested upstream sections whenassigning a minimum log TU of −4 and −5, respectively.Similar results were obtained for the ANOVAs, in which allsites in classes with a log TU > −3.5 exhibited a significantdifference (all p < 0.01) to reference sites (Figure 2).

The linear models exhibited the best-fit for the relationshipsbetween % kinvertebrate and SPEARpesticides, except for sites fromBrittany, France, for which a Weibull II model yielded a slightlybetter fit (Supporting Information Table S2). However, theslope for linear regression models was only significant for thesites from Victoria, Australia (r2 = 0.46, p < 0.001, n = 23) andBrittany, France (r2 = 0.87, p < 0.001, n = 11), but not for the

Figure 1. Dose−response curves with 95% confidence bands (gray) for the relationship between log TU and SPEARpesticides for sites with forestedupstream sections (a) and for sites without forested upstream sections (b). Reference sites were assigned a minimum log TU of −4. Random noise(0.00002) was added (jittering) to the TU values in order to show all data points in the plot. This affected primarily the sites with minimum TUs.

Figure 2. Arithmetic mean of SPEARpesticides values with standarderrors for different classes of toxic units (TU). The sampling sites weredivided into sites with forested upstream sections (filled points) andsites without forested upstream sections (open points). Sample sizes ofthe different classes were 44 and 9 sites with a log TU ≤ −3.5, 10, and12 sites with −3.5 < log TU ≤ −2.5, 12, and 11 sites with −2.5 < logTU ≤ −1.5 and 4 and 9 sites with log TU > −1.5 for sites withforested upstream sections and sites without forested upstreamsections, respectively. All classes with TUs > −3.5 were significantlydifferent (all p < 0.001) to reference sites in ANOVA with treatmentcontrasts.

Environmental Science & Technology Article

dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−51425137

sites from Denmark (r2 = 0.09, p = 0.31, n = 13), for which noplausible relationship between SPEARpesticides and %kinvertebratecould be established (data not shown). The sites fromDenmark were not included in the joint dose−responsemodeling, because the aim was to derive a joint relationshipbetween SPEARpesticides and %kinvertebrate. The best-fit model forthe joint data from France and Australia was linear (SupportingInformation Table S2) and exhibited a good fit between %kinvertebrate and SPEARpesticides (r

2 = 0.51, p < 0.001, n = 34)(Figure 3). In ANCOVA, the intercepts for data from Brittany

and Victoria were significantly different (p = 0.02), whereas theslopes exhibited no significant differences (p = 0.25). However,if the data of each region were autoscaled before ANCOVA,neither the intercepts (p = 0.89) nor slopes (p = 0.15) weresignificantly different.

■ DISCUSSIONEffect Thresholds for Macroinvertebrate Commun-

ities. In our analysis, pesticide effects on the abundance ofsensitive invertebrates were found at TUs for D. magna below0.01. Concentrations related to the safety factor incorporated inthe UP resulted in a 27% to 61% decline in the abundance ofsensitive taxa, depending on the presence of forested upstreamsections and which minimum TU was selected in modeling(Figure 1, Figure S1). Similarly, both of the latter factors(presence of forested upstream sections and minimum TU)influenced the estimated effect concentrations (Table 2) andthe concentration at which significant differences to referencesites occurred. The models with a lower minimum TUexhibited a better fit in terms of the BIC compared to modelswith a higher minimum TU (Table S2). Nevertheless, morefield data in the log TU range of −3 to −5 would be needed tosubstantiate a selection between both minimum TUs. Theeffect threshold was determined to be approximately 1 to 1.5orders of magnitude lower (log TU of −3 to −3.6) than thesafety factor of the UP. However, field studies and results fromthe joint analysis of studies with differing methodologies are

subject to random and systematic uncertainties that can lead towider confidence bands or bias in the dose−response modelsand would consequently affect the derived effect threshold.Several uncertainties were identified as random (SupportingInformation Table S4) and presumably resulted in widerconfidence bands of the fitted dose−response curves (Figure 1,Figure S1) as for example, the joint analysis of data obtainedfrom different countries, years and pesticide sampling methods(Table1). Nevertheless, there were no significant differencesbetween the studies regarding the relationship of SPEARpesticidesand TU (linear model for studies with pesticide gradients of atleast 1 log unit in terms of TU, p = 0.23, n = 58). Moreover,some of the variation in the relationship between pesticidetoxicity and SPEARpesticides may result from differences in thedose−response relationship of individual compounds i.e.concentrations of different compounds relating to for example1/100 of their EC50 for D. magna may exert different effects onthe abundance of sensitive taxa. Two studies found that thetoxicity of a range of different organic toxicants also explainedbetween 68% and 87% of the variance in terms of r2 inSPEAR,24,26,36 suggesting that the use of toxic units for D.magna as benchmark for the toxicity of different organictoxicants is adequate and that the associated uncertainty is ofminor importance.Two sources of uncertainty could result in a systematic bias

in the derived effect threshold. First, the underestimation ofpesticide toxicity due to underestimated pesticide concen-trations or the nonmeasurement of ecotoxicologically relevantcompounds would lead to a left shift of the dose−responsecurve and consequently a decrease in the effect threshold.However, underestimation of the real concentrations by a factorof 10 to 100 would be required in order to yield similar effectthresholds than incorporated in the UP. We consider anunderestimation of this order of magnitude as highly unlikelygiven that the field studies employed sampling techniquesespecially targeted at capturing episodic pesticide exposures(Table 1). Although the sampling techniques varied due todifferences in the pesticides monitored and regional conditions,the relationship between pesticide toxicity in terms of TU andSPEARpesticides was not significantly different between thestudies (see above). Furthermore, the concentrations respon-sible for the estimated TUs are in agreement with studiesmonitoring pesticide runoff with 15-min or 1-h resolution insingle agricultural streams. Two studies in Central Europefound peak herbicide concentrations of 2.537 and 3.5 μg/L,38,39

and a study in South Africa found insecticide concentrationsranging from 0.2 to 2.9 μg/L in runoff events.40 In the studieswith relevant pesticide contamination (TU > −4) included inour meta-analysis, peak insecticide concentrations were in therange between 0.3 and 1.2 μg/L (i.e., Brittany, France = 0.7 μg/L; Victoria, Australia = 1.2 μg/L; Denmark = 0.3 μg/L;Germany = 0.5 μg/L). Moreover, it would not be expected thatall compounds are underestimated in equal measure andunderestimation should therefore increase the variability inpesticide toxicity that relied on different compounds in allincluded studies. In fact, all individual studies exhibited a verygood fit (all r2 from linear models between 0.62 and 0.68) ofpesticide toxicity with the respective biotic end point. Finally,laboratory toxicity experiments with single pyrethoid, organo-chlorine and organophosphate insecticides demonstrated thatchronic population effects can occur 3 to 4 orders of magnitudebelow acute toxicity concentrations.41−44 This means thateffects may be expected above a TU for D. magna of −4 or −3,

Figure 3. Linear regression model for the relationship betweenSPEARpesticides and %kinvertebrate and for sampling sites from Brittany,France and Victoria, Australia. The linear model explained 51% of thevariation (p < 0.001, n = 34). Note that the intercepts for data fromBrittany and Victoria were significantly different (p = 0.02) inANCOVA (see Results for details).

Environmental Science & Technology Article

dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−51425138

which is in line with the effect thresholds derived here. Overall,although we agree that methodical advancements in thecharacterization of episodic pesticide exposure are desirableand that underestimation of true peak concentrations to acertain degree may occur in the field with currently availablesampling techniques, we consider it highly unlikely that truepesticide concentrations were 1 or 2 orders of magnitudehigher.Second, the response of the SPEAR index to a confounding

factor that is highly correlated with pesticide toxicity would leadto a decrease in effect thresholds (Supporting InformationTable S4). The original studies included in our analysis (Table1) and two recent studies45,46 identified the toxicity of theobserved stressor as the most important explanatory variablefor the respective version of the SPEAR index, whereas 8 and 9respectively measured confounding factors exhibited noexplanatory power for SPEAR. Furthermore, three studiesdemonstrated that SPEARpesticides decreased only in contami-nated sites and after beginning of the pesticide applicationperiod,7,17,47 reinforcing that pesticides are the culprit becauseother agricultural stressors such as eutrophication orsedimentation are present throughout the year. Moreover,SPEARpesticides showed no response to physicochemical, habitator landscape variables in reference sites.18,48 Finally, albeit onestudy found a difference in SPEARpesticides between samplings ofheterogeneous and homogeneous habitats in 13 streams inagricultural areas, there was nevertheless a strong relationshipbetween SPEARpesticides and TU for each habitat type (r2 = 0.68and 0.6 in heterogeneous and homogeneous habitats,respectively).49 Overall, there is only low uncertainty that theindicated effects were not due to pesticide toxicity. Higheruncertainty remains regarding the mechanisms causing theobserved effects of pesticides in the field. A review of mesocosmstudies regarding effects of carbamate, pyrethroid and organo-phosphate insecticides suggested that a log TU for D. magna of−2 would be protective for individual insecticides in the field.3

Beside methodical46 (and see debate50,51) reasons, thedifference between the field and mesocosm studies could resultfrom different community composition,52 repeated exposures,pesticide mixtures5,53 and the joint effects of differentstressors,4,46,54,55 all of which can enhance the effects ofpesticides and are rarely considered in mesocosms. Moreover,chronic long-term effects on merolimnic insects occurring atconcentrations related to a TU of −3 to −4 as outlinedabove41−44 may not be detected in mesocosm studies, whichrarely exceed a study period of several months. However,studies with a high temporal and spatial resolution would beneeded to clarify the mechanisms in the field [see ref 56].Overall, we suggest that there is low uncertainty that ourderived effect threshold for effects of pesticides on macro-invertebrate communities is too low and we therefore concludethat the safety factor related to D. magna incorporated in theEU Uniform Principles for single pesticides is not protective forfreshwater ecosystems, though the mechanisms should beelucidated in future studies.Effects Thresholds for the Ecosystem Function of Leaf

Breakdown. The relationship between the communitystructure in terms of SPEARpesticides and the percentage ofinvertebrate leaf breakdown was linear in Brittany, France andVictoria, Australia (Figure 3). This means that of severalsuggested links between the community structure andecosystem functions (see Introduction), pesticide effects onthe abundance of sensitive macroinvertebrates seem to translate

to a similar effect on the breakdown rate of leafs byinvertebrates. Hence, in these regions is no greater toleranceof this important ecosystem function to pesticide contami-nation and the effect thresholds for the abundance of SPEARtaxa may also apply. Given that regional case studies on therelationship between pesticides and ecosystem functions arescarce, this result may not hold for different ecosystemfunctions,8 and other regions. In fact, no plausible relationshipbetween SPEARpesticides and the invertebrate leaf breakdownrate was found for the Danish sites. Similarly, the original studyon the Danish streams only reported a statistically significantrelationship between pesticide toxicity and microbial leafbreakdown but not with invertebrate leaf breakdown.23

Hence, although pesticide toxicity lead to community changein terms of SPEARpesticides in the Danish sites (Figure 1), thisdid not translate to effects on the ecosystem function of leafbreakdown. This can be explained by the domination of theshredder community by Gammarus pulex, which is a rathertolerant species due to its ecological traits and is consequentlynot classified as SPEAR.23 In fact, the density of Gammaruspulex was significantly correlated only to the leaf breakdownrate (p = 0.02, test for Pearson correlation, n = 13). Thus, theeffect threshold for the invertebrate leaf breakdown presumablydepends on the composition of the shredder community, and ifnon-SPEAR taxa such as Gammarus pulex dominate, there maybe functional redundancy up to a certain threshold beforepesticides affect invertebrate leaf breakdown. Finally, thequestion is to which extent a temporal difference betweenpesticide application and leaf input from deciduous trees affectsthe relationship between pesticide-driven structural changesand invertebrate leaf breakdown (see ref 57). The studies inFrance and Denmark were conducted in the period of peakinsecticide application in these regions, which precedes theperiod of main input of leafs (late autumn) by several months.Although it is known that community alterations can persistover months,7,20 it remains to be shown that the invertebrateleaf breakdown is affected outside of the main season ofpesticide application. However, for streams receiving a relativelyconstant leaf input from evergreen forests (e.g., Australianstreams),58 the influence of seasonality on the effects ofpesticides on invertebrate leaf breakdown should be of minorimportance.

Relevance for Ecological Risk Assessment of AquaticEcosystems. The thresholds obtained in our study may berelevant for pesticides and other organic compounds wheremacroinvertebrates represent the most sensitive group of taxa.In a study on the concentrations of 331 organic toxicants inlarge rivers of North Germany, invertebrates were considered asmost sensitive for 110 compounds, among them manyinsecticides and fungicides, whereas algae and fish representedthe most sensitive group for 142 and 79 compounds,respectively.59 Another study reported that invertebrates weremost sensitive for 225 organic toxicants, whereas algae and fishexhibited highest sensitivity for 158 and 104 organic toxicants,respectively.36 In this study, an effect threshold of 1/1000 of theacute EC50 for D. magna was suggested for the derivation ofenvironmental quality standards (EQS) for river basin specificpollutants, based on an analysis of macroinvertebratebiomonitoring and chemical monitoring data. Hence, effectthresholds for macroinvertebrates would also be protective ofother aquatic organisms for a wide range of compounds. Therelatively good relationship between pesticide toxicity in termsof TU and SPEAR in our study is remarkable, considering that

Environmental Science & Technology Article

dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−51425139

the macroinvertebrate data originated from different regions inEurope, Siberia and Australia and the low explanatory power forbiotic end points often seen in ecological meta-analyses.60 Ourstudy therefore supports the use of trait-based approaches inrisk assessment to identify the impact of anthropogenicstressors on a continental or even global scale.11,16

Without knowing the temporal and spatial dimension of thereduction in the abundance of sensitive macroinvertebratepopulations in this study, it is not possible to decide whetherthe observed effects on the communities were transient or long-term, defined as no complete recovery until the spraying periodin the consecutive year. In the latter case, the effects would beunacceptable for the requirements of the EU directive for theplacement of plant protection products on the market.2,61

However, we suggest that current exposure of freshwaterecosystems to pesticides may be unacceptable for therequirements of this and other EU directives. First, one fieldstudy showed long-term effects, that is, that no recovery of thecommunities occurred until the prespraying period of thefollowing year.7 Second, given that pesticides are widely appliedin agriculture, which represents the dominant land use in theEU and elsewhere, and that pesticides frequently occur instreams and rivers in concentrations above effect thresholds,20

the associated reduction in the abundance of sensitive taxa maylead to losses in biodiversity on a regional scale (γ-diversity) asalso indicated by other studies.62,63 Since a recent EU Directiverequires that the risks for biodiversity from pesticides beminimized,64 more pesticide mitigation measures may beneeded to comply with this Directive. Our study highlightedon the basis of a comprehensive data set that forested upstreamsections can reduce adverse effects of pesticides on themacroinvertebrate community, especially under low pesticidecontamination as indicated by higher effect thresholds (Table2). Hence, together with other risk mitigation measures such aspesticide use reduction, buffer strips and vegetated treatmentsystems,22,65,66 the conservation and increase of landscapepatches without agricultural disturbance may somewhatalleviate the effects of pesticides in aquatic ecosystems.

■ ASSOCIATED CONTENT

*S Supporting InformationA figure with the relationship between pesticide toxicity andSPEARpesticides for a minimum log TU of −5, a table with thegoodness of fit results for all models, a table with theparameters of the fitted dose−response models and a tablegiving sources of potential uncertainties. This material isavailable free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATION

Corresponding Author*Telephone: ++49 (0) 6341 28031536. E-mail: senator@ecotoxicology.de.

Author ContributionsStudy design: R.B.S., P.C.O., B.J.K., M.B., M.L. Provision ofdata: R.B.S., P.C.O., J.R., M.B., M.L. Data analysis: R.B.S.Discussion and interpretation of results: all. Drafting ofmanuscript: R.B.S. Revising manuscript: R.B.S., M.B., B.J.K.,P.C.O., M.L.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTS

The authors thank all persons involved in the field studiesanalysed in this paper. Mirco Bundschuh and five anonymousreviewers gave valuable comments that improved the quality ofthe manuscript. We are grateful to Sabine Duquesne forclarifying the legal situation for pesticide authorization. P.C.O.was financially supported through a Deutsche Forschungsge-meinschaft (DFG) postdoctoral fellowship (PAK 406/1). J.R.received funding from the Danish Research Council (RISK-POINT, Grant No 2104-07-0035). A visit by B.J.K. to Germanywas funded by the University of Technology Sydney.

■ REFERENCES(1) Millenium Ecosystem Assessment Ecosystems and Human Well-being: Synthesis; Island Press: Washington, DC, 2005.(2) European Council Directive 91/414/EEC of 15 July 1991concerning the placing of plant protection products on the market.(3) Van Wijngaarden, R. P. A; Brock, T. C. M.; Van Den Brink, P. J.Threshold levels for effects of insecticides in freshwater ecosystems: Areview. Ecotoxicology 2005, 14 (3), 355−380.(4) Belden, J. B.; Gilliom, R. J.; Martin, J. D.; Lydy, M. J. Relativetoxicity and occurrence patterns of pesticide mixtures in streamsdraining agricultural. Integr. Environ. Assess. Manage. 2007, 3 (1), 90−100.(5) Van den Brink, P. J.; Crum, S. J. H.; Gylstra, R.; Bransen, F.;Cuppen, J. G. M.; Brock, T. C. M. Effects of a herbicide-insecticidemixture in freshwater microcosms: Risk assessment and ecologicaleffect chain. Environ. Pollut. 2009, 157 (1), 237−249.(6) Townsend, C. R.; Uhlmann, S. S.; Matthaei, C. D. Individual andcombined responses of stream ecosystems to multiple stressors. J.Appl. Ecol. 2008, 45 (6), 1810−1819.(7) Liess, M.; von der Ohe, P. C. Analyzing effects of pesticides oninvertebrate communities in streams. Environ. Toxicol. Chem. 2005, 24(4), 954−965.(8) Schafer, R. B.; Bundschuh, M.; Rouch, D. A.; Szocs, E.; von derOhe, P. C.; Pettigrove, V.; Schulz, R.; Nugegoda, D.; Kefford, B. J.Effects of pesticide toxicity, salinity, and other environmental variableson selected ecosystem functions in streams and the relevance forecosystem services. Sci. Total Environ. 2012, 415 (1), 69−78.(9) Webster, J. R. Spiraling down the river continuum: streamecology and the U-shaped curve. J. North Am. Benthol. Soc. 2007, 26(3), 375−389.(10) Sutherland, W. J.; Adams, W. M.; Aronson, R. B.; Aveling, R.;Blackburn, T. M.; Broad, S.; Ceballos, G.; Cote, I. M.; Cowling, R. M.;GA, D. A. F.; Dinerstein, E.; Ferraro, P. J.; Fleishman, E.; Gascon, C.;Hunter, M., Jr.; Hutton, J.; Kareiva, P.; Kuria, A.; Macdonald, D. W.;Mackinnon, K.; Madgwick, F. J.; Mascia, M. B.; McNeely, J.; Milner-Gulland, E. J.; Moon, S.; Morley, C. G.; Nelson, S.; Osborn, D.; Pai,M.; Parsons, E. C.; Peck, L. S.; Possingham, H.; Prior, S. V.; Pullin, A.S.; Rands, M. R.; Ranganathan, J.; Redford, K. H.; Rodriguez, J. P.;Seymour, F.; Sobel, J.; Sodhi, N. S.; Stott, A.; Vance-Borland, K.;Watkinson, A. R. One hundred questions of importance to theconservation of global biological diversity. Conserv. Biol. 2009, 23 (3),557−567.(11) Cadotte, M. W.; Carscadden, K.; Mirotchnick, N. Beyondspecies: Functional diversity and the maintenance of ecologicalprocesses and services. J. Appl. Ecol. 2011, 48 (5), 1079−1087.(12) Kefford, B. J.; Schafer, R. B.; Metzeling, L. Risk assessment ofsalinity and turbidity in Victoria (Australia) to stream insects’community structure does not always protect functional traits. Sci.Total Environ. 2012, 415 (1), 61−68.(13) Van den Brink, P.; Alexander, A. C.; Desrosiers, M.; Goedkoop,W.; Goethals, P. L. M.; Liess, M.; Dyer, S. D. Traits-based approachesin bioassessment and ecological risk assessment: Strengths, weak-nesses, opportunities and threats. Integr. Environ. Assess. Manage. 2011,7 (2), 198−208.

Environmental Science & Technology Article

dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−51425140

(14) Liess, M.; Schafer, R. B.; Schriever, C. A. The footprint ofpesticide stress in communitiesSpecies traits reveal communityeffects of toxicants. Sci. Total Environ. 2008, 406 (3), 484−490.(15) Poff, N. L. Landscape filters and species traits: Towardsmechanistic understanding and prediction in stream ecology. J. NorthAm. Benthol. Soc. 1997, 16 (2), 391−409.(16) Statzner, B.; Beche, L. A. Can biological invertebrate traitsresolve effects of multiple stressors on running water ecosystems?Freshwat. Biol. 2010, 55, 80−119.(17) Schafer, R. B.; Caquet, T.; Siimes, K.; Mueller, R.; Lagadic, L.;Liess, M. Effects of pesticides on community structure and ecosystemfunctions in agricultural streams of three biogeographical regions inEurope. Sci. Total Environ. 2007, 382 (2−3), 272−285.(18) von der Ohe, P. C.; Pruß, A.; Schafer, R. B.; Liess, M.; Deckere,E. d.; Brack, W. Water quality indices across Europe - a comparison ofthe good ecological status of five river basins. J. Environ. Monit. 2007,9, 970−978.(19) Schafer, R. B.; Pettigrove, V.; Rose, G.; Allinson, G.; Wightwick,A.; von der Ohe, P. C.; Shimeta, J.; Kuhne, R.; Kefford, B. J. Effects ofpesticides monitored with three sampling methods in 24 sites onmacroinvertebrates and microorganisms. Environ. Sci. Technol. 2011,45 (4), 1665−1672.(20) Schafer, R. B.; van den Brink, P. J.; Liess, M. Impacts ofpesticides on freshwater ecosystems. In Ecological Impacts of ToxicChemicals; Sanchez-Bayo, F., van den Brink, P., Mann, R. M., Eds.;Bentham: Bussum, NL, 2011.(21) Friberg, N.; Lindstrom, M.; Kronvang, B.; Larsen, S. E.Macroinvertebrate/sediment relationships along a pesticide gradient inDanish streams. Hydrobiologia 2003, 494 (1−3), 103−110.(22) Rasmussen, J. J.; Baattrup-Pedersen, A.; Wiberg-Larsen, P.;McKnight, U. S.; Kronvang, B. Buffer strip width and agriculturalpesticide contamination in Danish lowland streams: Implications forstream and riparian management. Ecol. Eng. 2011, 37 (12), 1990−1997.(23) Rasmussen, J. J.; Wiberg-Larsen, P.; Baattrup-Pedersen, A.;Monberg, R. J.; Kronvang, B. Impact of pesticides and natural stressorson leaf litter decomposition in agricultural streams. Sci. Total Environ.2011, 416, 148−155.(24) Beketov, M. A.; Liess, M. An indicator for effects of organictoxicants on lotic invertebrate communities: Independence ofconfounding environmental factors over an extensive river continuum.Environ. Pollut. 2008, 156 (3), 980−987.(25) Sprague, J. B. Measurement of pollutant toxicity to fish, II-Utilizing and applying bioassay results. Water Res. 1970, 4 (1), 3−32.(26) von der Ohe, P. C.; de Deckere, E.; Pruß, A.; Munoz, I.;Wolfram, G.; Villagrasa, M.; Ginebreda, A.; Hein, M.; Brack, W.Toward an integrated assessment of the ecological and chemical statusof European river basins. Integr. Environ. Assess. Manage. 2009, 5 (1),50−61.(27) Beketov, M. A.; Foit, K.; Schafer, R. B.; Schriever, C. A.; Sacchi,A.; Capri, E.; Biggs, J.; Wells, C.; Liess, M. SPEAR indicates pesticideeffects in streamsComparative use of species- and family-levelbiomonitoring data. Environ. Pollut. 2009, 157, 1841−1848.(28) Benfield, E. F. Decomposition of leaf material. In Methods inStream Ecology; Hauer, F. R.; Lamberti, G. A., Eds.; Academic Press:San Diego, CA, 2007.(29) Gessner, M. O.; Chauvet, E. A case for using litter breakdown toassess functional stream integrity. Ecol. Appl. 2002, 12 (2), 498−510.(30) Statzner, B.; Bis, B.; Doledec, S.; Usseglio-Polatera, P.Perspectives for biomonitoring at large spatial scales: A unifiedmeasure for the functional composition of invertebrate communities inEuropean running waters. Basic Appl. Ecol. 2001, 2, 73−85.(31) Allan, J. D.; Castillo, M. M., Stream Ecology: Structure andFunction of Running Waters. Springer: Dordrecht, the Netherlands,2007.(32) Ritz, C. Toward a unified approach to dose-response modelingin ecotoxicology. Environ. Toxicol. Chem. 2010, 29 (1), 220−229.(33) Fox, J.; Weisberg, S. An R Companion to Applied Regression, 2nded.; Sage: Los Angeles, CA, 2011.

(34) R Development Core Team R: A language and environment forstatistical computing, reference index version 2.13.1. www.r-project.org(October 5 2011).(35) Ritz, C.; Streibig, J. C. Bioassay analysis using R. J. Stat. Software2005, 12 (5), 1−22.(36) von der Ohe, P. C.; Dulio, V.; Slobodnik, J.; De Deckere, E.;Kuhne, R.; Ebert, R.-U.; Ginebreda, A.; De Cooman, W.; Schuurmann,G.; Brack, W. A new risk assessment approach for the prioritization of500 classical and emerging organic microcontaminants as potentialriver basin specific pollutants under the European Water FrameworkDirective. Sci. Total Environ. 2011, 409 (11), 2064−2077.(37) Wittmer, I. K.; Bader, H. P.; Scheidegger, R.; Singer, H.; Luck,A.; Hanke, I.; Carlsson, C.; Stamm, C. Significance of urban andagricultural land use for biocide and pesticide dynamics in surfacewaters. Water Res. 2010, 44 (9), 2850−2862.(38) Leu, C.; Singer, H.; Stamm, C.; Muller, S. R.; Schwarzenbach, R.P. Simultaneous assessment of sources, processes, and factorsinfluencing herbicide losses to surface waters in a small agriculturalcatchment. Environ. Sci. Technol. 2004, 38 (14), 3827−3834.(39) Leu, C.; Singer, H.; Stamm, C.; Muller, S. R.; Schwarzenbach, R.P. Variability of herbicide losses from 13 fields to surface water withina small catchment after a controlled herbicide application. Environ. Sci.Technol. 2004, 38 (14), 3835−3841.(40) Schulz, R. Rainfall-induced sediment and pesticide input fromorchards into the Lourens River, Western Cape, South Africa:Importance of a single event. Water Res. 2001, 35, 1869−1876.(41) Beketov, M. A.; Liess, M. Acute contamination withesfenvalerate and food limitation: Chronic effects on the mayfly,Cloeon dipterum. Environ. Toxicol. Chem. 2005, 24 (5), 1281−1286.(42) Liess, M. Population response to toxicants is altered byintraspecific interaction. Environ. Toxicol. Chem. 2002, 21 (1), 138−142.(43) Liess, M.; Schulz, R. Chronic effects of short-termcontamination with the pyrethroid insecticide fenvalerate on thecaddisfly Limnephilus lunatus. Hydrobiologia 1996, 324 (2), 99−106.(44) Schulz, R.; Liess, M. Chronic effects of low insecticideconcentrations on freshwater caddisfly larvae. Hydrobiologia 1995,299, 103−113.(45) Schafer, R. B.; Kefford, B.; Metzeling, L.; Liess, M.; Burgert, S.;Marchant, R.; Pettigrove, V.; Goonan, P.; Nugegoda, D. A traitdatabase of stream invertebrates for the ecological risk assessment ofsingle and combined effects of salinity and pesticides in South-EastAustralia. Sci. Total Environ. 2011, 409 (11), 2055−2063.(46) Liess, M.; Beketov, M. Traits and stress: keys to identifycommunity effects of low levels ot toxicants in test systems.Ecotoxicology 2011, 20 (6), 1328−1340.(47) Schriever, C. A.; Hansler-Ball, M.; Holmes, C.; Maund, S.; Liess,M. Agricultural intensity and landscape structure: influences on themacroinvertebrate assemblages of small streams in northern Germany.Environ. Toxicol. Chem. 2007, 26 (2), 346−357.(48) Schletterer, M.; Fureder, L.; Kuzovlev, V. V.; Beketov, M. A.Testing the coherence of several macroinvertebrate indices andenvironmental factors in a large lowland river system (Volga River,Russia). Ecol. Indic. 2010, 10 (6), 1083−1092.(49) Rasmussen, J. J.; Wiberg-Larsen, P.; Baattrup-Pedersen, A.;Friberg, N.; Kronvang, B. Stream habitat structure influencesmacroinvertebrate response to pesticides. Environ. Pollut. 2012, 164,142−149.(50) Van den Brink, P. J.; Ter Braak, C. J. Response to “Traits andstress: keys to identify community effects of low levels of toxicants intest systems” by Liess and Beketov (2011). Ecotoxicology 2012, 21 (2),297−9.(51) Liess, M.; Beketov, M. A. Rebuttal related to “Traits and stress:Keys to identify community effects of low levels of toxicants in testsystems” by Liess and Beketov (2011). Ecotoxicology 2012, 21 (2),300−3.(52) Beketov, M.; Schafer, R. B.; Marwitz, A.; Paschke, A.; Liess, M.Long-term stream invertebrate community alterations induced by the

Environmental Science & Technology Article

dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−51425141

insecticide thiacloprid: Effect concentrations and recovery dynamics.Sci. Total Environ. 2008, 405, 96−108.(53) Lydy, M. J.; Austin, K. R. Toxicity assessment of pesticidemixtures typical of the Sacramento-San Joaquin Delta usingChironomus tentans. Arch. Environ. Contam. Toxicol. 2005, 48 (1),49−55.(54) Holmstrup, M.; Bindesbol, A. M.; Oostingh, G. J.; Duschl, A.;Scheil, V.; Kohler, H. R.; Loureiro, S.; Soares, A. M.; Ferreira, A. L.;Kienle, C.; Gerhardt, A.; Laskowski, R.; Kramarz, P. E.; Bayley, M.;Svendsen, C.; Spurgeon, D. J. Interactions between effects ofenvironmental chemicals and natural stressors: A review. Sci. TotalEnviron. 2010, 408 (18), 3746−62.(55) Heugens, E.; Hendriks, A.; Dekker., T.; Van Straalen, N. M.;Admiraal, W. A review of the effects of multiple stressors on aquaticorganisms and analysis of uncertainty factors for use in risk assessment.Crit. Rev. Toxicol. 2001, 31 (3), 247−284.(56) Liess, M.; Schulz, R. Linking insecticide contamination andpopulation response in an agricultural stream. Environ. Toxicol. Chem.1999, 18, 1948−1955.(57) Bergfur, J. Seasonal variation in leaf-litter breakdown in nineboreal streams: implications for assessing functional integrity. Fundam.Appl. Limnol. 2007, 169 (4), 319−329.(58) Tyler, P.; Williams, B. Lakes and Rivers. In Ecology: AnAustralian Perspective; Attiwill, P., Wilson, B., Eds.; Oxford UniversityPress: South Melbourne, Australia, 2003.(59) Schafer, R. B.; von der Ohe, P.; Kuhne, R.; Schuurmann, G.;Liess, M. Occurrence of 331 organic pollutants in four rivers of NorthGermany between 1994 and 2004 and risk assessment for algae,invertebrates and fish. Environ. Sci. Technol. 2011, 45 (14), 6167−6174.(60) Moller, A. P.; Jennions, M. D. How much variance can beexplained by ecologists and evolutionary biologists? Oecologia 2002,132 (4), 492−500.(61) European Commission Guidance Document Sanco/3268/2001rev.4 on Aquatic Ecotoxicology in the context of the Directive 91/414/EEC..(62) Gibbs, K. E.; Mackey, R. L.; Currie, D. J. Human land use,agriculture, pesticides and losses of imperiled species. Divers. Distrib.2009, 15 (2), 242−253.(63) Geiger, F.; Bengtsson, J.; Berendse, F.; Weisser, W. W.;Emmerson, M.; Morales, M. B.; Ceryngier, P.; Liira, J.; Tscharntke, T.;Winqvist, C.; Eggers, S.; Bommarco, R.; Part, T.; Bretagnolle, V.;Plantegenest, M.; Clement, L. W.; Dennis, C.; Palmer, C.; Onate, J. J.;Guerrero, I.; Hawro, V.; Aavik, T.; Thies, C.; Flohre, A.; Hanke, S.;Fischer, C.; Goedhart, P. W.; Inchausti, P. Persistent negative effects ofpesticides on biodiversity and biological control potential on Europeanfarmland. Basic Appl. Ecol. 2010, 11 (2), 97−105.(64) European Commission Directive 2009/128/EC of the EuropeanParliament and of the Council of 21 October 2009 establishing aframework for Community action to achieve the sustainable use ofpesticides.(65) Reichenberger, S.; Bach, M.; Skitschak, A.; Frede, H. G.Mitigation strategies to reduce pesticide inputs into ground- andsurface water and their effectiveness; A review. Sci. Total Environ. 2007,384 (1−3), 1−35.(66) Stehle, S.; Elsaesser, D.; Gregoire, C.; Imfeld, G.; Passeport, E.;Payraudeau, S.; Schafer, R. B.; Tournebize, J.; Schulz, R. Pesticide RiskMitigation by Vegetated Treatment Systems: A Meta-analysis. J.Environ. Qual. 2011, 40 (4), 1068−1080.

Environmental Science & Technology Article

dx.doi.org/10.1021/es2039882 | Environ. Sci. Technol. 2012, 46, 5134−51425142